Fri, 7 Sep 90 00:48:08 O
|
Well I couldn't reproduce this problem, but I know how to solve it: I
will remove extended statistics from the next release, as they have
basically become useless:
1. For lists with a non-negligible number of non-local recipients and
non-negligible traffic, you should use DISTRIBUTE and in this case
such statistics are not computed.
2. If the traffic is very low (like LCOORD-L@CEARN) you don't need to
have extended statistics.
3. For lists with only local people you should not use DISTRIBUTE but the
link.bytes value will be 0 since you are not sending anything outside
the node.
Extended stats cost a lot of CPU time to maintain and have become
worthless through the introduction of domain-style addresses which cannot
be used in the calculation. I would not be surprised if a lot of list
owners used the 'Extended' option just because they copied it from
another list, without understanding, or because they just want "more
statistics" (ie they don't know what it means, but it means "more" and
they want as much as they can). Let's save cycles and prevent people from
drawing wrong conclusions from meaningless figures.
Eric
|
|
|